Video Analysis Watch Table for Forensic Alert Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video analysis systems for forensic purposes face delays and missed alerts due to requiring a defined threshold for generating alarms, which can lead to false negatives when the quality of extracted images is insufficient, causing high-priority watch-list entries to be missed.
Innovation Solution
A computer-implemented method that records and analyzes video data by extracting identification information, computing matching values, and storing object data in a watch table, allowing for real-time sorting and updating based on similarity, with the ability to display metadata and images, and maintaining a limited list of highest matching entries for efficient tracking and alert generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a defined threshold is used to generate alarms, then false positives are reduced, but high-priority matches may be missed when image quality is insufficient
Solution Approach 1:
The system performs preliminary extraction and storage of identification information from video frames before final matching. By pre-processing and storing extracted features with their corresponding matching values, the system preserves all potential matches regardless of threshold, allowing post-processing review of high-priority cases that initially fell below the alarm threshold.
Solution Approach 2:
The system processes and stores more matching results than traditionally required by the threshold. Instead of discarding sub-threshold matches, it retains them in the data structure for potential review, performing excessive processing to ensure no high-priority matches are lost, with the understanding that not all stored matches will be actionable.
2Adaptability or versatility
If all best shots are stored for reverse search, then comprehensive analysis is enabled, but storage requirements and processing time increase
Solution Approach 1:
The system extracts only the essential identification information and matching values from video frames, storing them in a structured data format. By separating the critical matching data from the complete video footage, the system enables reverse search functionality with significantly reduced storage requirements while maintaining the capability to perform comprehensive analysis when needed.
Solution Approach 2:
The system creates a simplified copy of the video data in the form of extracted identification information and matching values. This compressed representation preserves the essential matching capabilities for reverse search without requiring storage of the entire video stream, allowing efficient data retrieval and analysis.
3Speed
If real-time surveillance with threshold-based alarms is implemented, then immediate alerts are generated, but delays occur in searching and ranking recorded data
Solution Approach 1:
The system performs preliminary extraction and organization of identification information as video frames are captured, preparing the data structure in advance. This pre-processing allows the system to quickly retrieve and rank matches when a reverse search is initiated, eliminating the need to process and analyze raw video data from scratch during the search phase.
Solution Approach 2:
The system maintains a dynamic data structure that automatically updates with new matching results as video analysis continues. This dynamic organization allows efficient insertion, sorting, and retrieval of matches based on their matching values, enabling rapid response both for real-time threshold-based alarms and for subsequent reverse search operations.
Data Source
AI summary
The invention provides a method of analyzing video data comprising the steps of recording the video data using one or more video recording devices; analyzing at least one video frame of the video data to extract identification information regarding a specific object of interest that is comprised in a watch list of a plurality of objects of interest, the watch list including at least one entry for each of the plurality of objects of interest; computing a matching value by comparing the extracted identification information with predetermined identification information of the specific object of interest; and storing object data and the matching value as an additional entry in the watch list in association with the specific object of interest.